CN102016809A - Memory controller, memory system, semiconductor integrated circuit, and memory control method - Google Patents
Memory controller, memory system, semiconductor integrated circuit, and memory control method Download PDFInfo
- Publication number
- CN102016809A CN102016809A CN2009801141803A CN200980114180A CN102016809A CN 102016809 A CN102016809 A CN 102016809A CN 2009801141803 A CN2009801141803 A CN 2009801141803A CN 200980114180 A CN200980114180 A CN 200980114180A CN 102016809 A CN102016809 A CN 102016809A
- Authority
- CN
- China
- Prior art keywords
- mentioned
- access
- instruction
- data
- group
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/14—Handling requests for interconnection or transfer
- G06F13/16—Handling requests for interconnection or transfer for access to memory bus
- G06F13/1605—Handling requests for interconnection or transfer for access to memory bus based on arbitration
- G06F13/1642—Handling requests for interconnection or transfer for access to memory bus based on arbitration with request queuing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F12/00—Accessing, addressing or allocating within memory systems or architectures
- G06F12/02—Addressing or allocation; Relocation
- G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
- G06F12/0802—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
- G06F12/0862—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches with prefetch
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/40—Transformation of program code
- G06F8/41—Compilation
- G06F8/44—Encoding
- G06F8/443—Optimisation
- G06F8/4441—Reducing the execution time required by the program code
- G06F8/4442—Reducing the number of cache misses; Data prefetching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2212/00—Indexing scheme relating to accessing, addressing or allocation within memory systems or architectures
- G06F2212/60—Details of cache memory
- G06F2212/6022—Using a prefetch buffer or dedicated prefetch cache
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Dram (AREA)
Abstract
Disclosed is a memory controller (101) provided with: a command generator (102) that generates multiple access commands that include physical addresses, on the basis of address requests, including logical addresses, that indicate rectangular regions within image data; and a command issuer (105) that issues the multiple access commands generated by the command generator (102) to a memory (0). The command generator (102) has a group determiner (104) that determines to which group a bank that includes data to be accessed belongs to, on the basis of the physical address that corresponds to an access request. When data to be accessed consecutively straddles two banks belonging to different groups, a pair of a first and a second access command is generated, with said access commands sharing a prefetch buffer between the two banks belonging to different groups.
Description
Technical field
The present invention relates in image processing system, efficiently carry out storage control device, accumulator system, SIC (semiconductor integrated circuit) and the memory control methods that data transmit.
Background technology
In general, comprising in the signal conditioning package of civilian image processing apparatus, in order to store huge data, using high capacity and DRAM cheaply.Especially, in recent years image processing apparatus, owing at MPEG2 or H.264 waits the 3D graphics process etc. of the reply of HD (High Definition) Flame Image Process, multichannel processing simultaneously, high picture element, not only need to satisfy memory span, also need to have the DRAM of higher data transfer capability.In order to realize the higher data transfer capability, known following method: (1) improves the method for the operating frequency of bus, and (2) widen the method for the highway width of storer, (3) or and with the method for above-mentioned (1), (2) method.
On the other hand, when DRAM is carried out access, need to specify in advance to want the memory bank of access capable (bank row), activate processing.In addition, in same memory bank, when change institute access capable, need be to temporary transient access the line precharge of advancing handle, and to the line activating processing of advancing of new access.During these activate to be handled, during precharge handles, can not carry out access to corresponding memory bank, so in same memory bank, during switch line, generation can not access during, produce idle on the data bus.At this, in order to remedy this shortcoming, in general DRAM access control, transmit in the execution in the data that transmit data to certain specific memory body, handle by activation processing and the precharge of carrying out other memory banks, hidden above-mentioned can not access during, carry out and DRAM between data bus on can transmit control data, memory bank interleaving access (bank interleave) all the time.In order to make this memory bank interleaving access effectively work, need the continuous data delivery time of lengthening at same memory bank, and utilize to the data that memory bank beyond this memory bank transmits send carry out during can not access to certain memory bank above-mentioned hidden.
As cause during can not access at this the existing solution of the low problem of transmission efficiency, the method that has patent documentation 1 to be put down in writing.In the method, signal according to counter 0,1, with specific timing (timing) alternately timesharing access memory A, B, utilize thus the data transfer time of other storeies come hidden in single storer, produced can not access during, realize the raising of the access efficiency of bus thus.But in above-mentioned prior art, even owing to use a plurality of storeies, the storer of access simultaneously is 1, thus system the maximum memory band territory that can use be limited at 1 storer band territory that storer had.
Figure 1A is the figure of kind, operating frequency and burst (burst) length that DRAM of the prior art is shown.This figure relates to 4 kinds of DRAM, i.e. SDR (Single Data Rate) SDRAM (Synchronous DRAM), DDR (Double Data Rate) SDRAM, DDR2SDRAM, DDR3 SDRAM (below, abbreviate SDR, DDR, DDR2, DDR3 as).Illustrate these 4 DRAM internal bus operating frequency and data bus operating frequency separately.In addition, data bus has 32 bits, 64 and manys the highway width of bit than top grade, but in the figure for the purpose of simplifying the description, only shows the part corresponding with 1 bit.
SDR comprises memory core and input and output (I/O) impact damper.Memory core is corresponding to 1 memory cell array, via inputoutput buffer to the data of data bus input and output by 1 bit of the memory cell array of row address and column address appointment.The internal bus operating frequency (133MHz) of SDR is identical with external data bus operating frequency (133MHz).
Each memory core of DDR, DDR2, the DDR3 roughly memory core with SDR is identical.
The upper limit of the operating frequency of internal bus is that the upper limit according to the operating frequency of memory core decides.That is, the frequency of the upper limit that the capacitor that uses as memory cell can respond may be thought of as roughly 200MHz, and the operating frequency of memory core can not surpass roughly 200MHz yet.Be directed to this, the main frame that storer is carried out access is high speed year by year.In DDR, DDR2, DDR3, as making the method for DRAM high speed, the data of a plurality of bits of input and output concurrently between inputoutput buffer and memory core, inputoutput data serially between inputoutput buffer and data bus.
The inputoutput buffer of DDR and memory core between with the operating frequency of the 133MHz data of input and output 2 bits concurrently, and data bus between in fact with 2 times frequency 266MHz inputoutput data serially.Minimal burstiness pulse length among the DDR (being also referred to as basic burst length) becomes 2.
The inputoutput buffer of DDR2 and memory core between with the operating frequency of the 133MHz data of input and output 4 bits concurrently, and data bus between in fact with 4 times frequency 533MHz inputoutput data serially.Minimal burstiness pulse length among the DDR2 becomes 4.
The inputoutput buffer of DDR3 and memory core between with the operating frequency of the 133MHz data of input and output 8 bits concurrently, and data bus between in fact with 8 times frequency 1066MHz inputoutput data serially.Minimal burstiness pulse length among the DDR3 becomes 8.But in DDR3, for keep and DDR2 between compatibility, back half 4 bit in 8 bits that keep by abandoning in the inputoutput buffer come the function (burst sudden change (bust chop) function) of supported burst length 4.
Like this, SDRAM is for this problem of the high speed difficulty of the operating frequency that solves internal bus, and is many than specialization by inputoutput buffer is carried out, and realizes that the high speed of the operating frequency of external data bus is the expansion of memory area.
Figure 1B has supposed to make the effectively figure of an example of the data configuration of the situation of work of memory bank interleaving access in general DRAM.To be arranged in first memory bank, (at SDRAM is 1 with the continuous data of same row address and basic burst length, be 2 in DDR, being 4 in DDR2, is 8 etc. in DDR3) * data that are made of above-mentioned basic access unit that data (the being called basic access unit later on) back of the N byte that highway width is represented, configuration are arranged in second memory bank carry out.Same later on to being arranged in being configured of M memory bank with the continuous data of same row address.By carrying out like this under the situation of storage access, a plurality of memory banks evenly occur, so can efficiently implement the memory bank interleaving access.In Figure 1B, showing the memory bank number as an example is 2 situation.In addition, in general DRAM, can not carry out transmission (if the request primary access, then being bound to take place the input and output of the data of basic burst length * highway width), so basic access unit becomes identical with minimum access unit smaller or equal to above-mentioned basic burst length.
Fig. 2 illustrates at the example of the serial data generation of disposing as Figure 1B to the situation of the access of unwanted data.As shown in Figure 2, be requested in the data area that is requested access (back is called the access request zone) under the situation of N byte of centre of basic access unit, because minimum access unit is identical with basic access unit, so need be to comprising all basic access unit's request msgs in access request zone, the result need carry out the access of 2N byte (back is called the access essential regions) as shown in Figure 2.The result is at access essential regions 2N byte, and the access request zone just is the N byte, so become unwanted data as the amount of the N byte of its difference, data-transmission efficiency reduces.
Patent documentation 1: Japanese kokai publication hei 9-190376 communique
But, as the MPEG4 of the Image Compression of higher level or in H.264 waiting, with respect to existing Image Compression, in order when guaranteeing higher picture element, to realize low bit rate and high compression rate, be that the situation that decoding processing unit carries out decoding processing becomes many with 4 * 4,8 * 8 such small pixel units.For so little pixel is carried out decoding processing, need obtain littler pixel data from storer, but in order to realize higher data movement capacity, suppose the method for the highway width of enforcement extended memory, the data quantitative change that then once obtains is many, so the access at unwanted data becomes many, data-transmission efficiency reduces.In addition, even implement to improve the method for operating frequency, for example if bring up to the action frequency band of DDR3 from the action frequency band of DDR2, then the maximum band of system improves on the other hand, the minimal burstiness pulse length of 1 instruction unit increases, minimum data amount in 1 access increases, and the result increases at the access of unwanted data, and same data-transmission efficiency reduces.
But, in common DRAM, at primary access, though a memory bank being carried out the data of highway width * basic burst length amount transmits, but in the DRAM of further improvement, the memory bank that will be positioned at DRAM inside is divided into a plurality of groups (for example organizing A, B), will and certain organize after transmission between (for example organizing A) carries out half, transmission can be inserted, the access of the burst length of original half can be realized to its different group (for example organizing B).For example, be 8 in basic burst length, the memory bank number is among 8 the improvement DRAM, per 4 memory banks are divided into two groups A, B, after at first the data that the memory bank 0 of group A is carried out 4 bursts transmitted, the data that can carry out 4 bursts to the memory bank 0 (or 1,2,3) of group B transmitted.Afterwards, data transmission etc. can be carried out, the transmission shorter can be carried out than basic burst length at the memory bank 0 (or 1,2,3) of group A.But,, equally only considering in the control of memory bank with prior art in order effectively to use such improvement DRAM, data-transmission efficiency almost can not get improving, perhaps in the control of simple consideration group, also have at the continuous situation of same group transmission, data-transmission efficiency reduces.
Summary of the invention
The present invention solves above-mentioned prior art problems, and its purpose is to provide storage control device, accumulator system, SIC (semiconductor integrated circuit) and the memory control methods of the high efficiency that a kind of realization data of having used the improvement DRAM with the memory bank that is divided into a plurality of groups transmit.
In order to solve the problems of the technologies described above, storage control device of the present invention, be used to control access to storer, this storer possesses first group that comprises a plurality of memory banks, comprise second group of a plurality of memory banks, the prefetch buffer of N bit, wherein N is the integer more than 2, this memory stores view data also is carried out the burst access by burst length N, above-mentioned storage control device comprises: the instruction generating unit, access request according to the logical address that comprises the rectangular area in the above-mentioned view data of expression generates a plurality of access instructions that comprise physical address; And portion is sent in instruction, send the above-mentioned a plurality of access instructions that generate by the instruction generating unit to above-mentioned storer, above-mentioned instruction generating unit has the group judging part, this group judging part basis and above-mentioned access request physical address corresponding, judge which group the memory bank that comprises the data of wanting access belongs to, stride two memory banks of the group that belongs to different in the data of wanting access and consecutive hours, generation comprises above-mentioned a plurality of access instructions of first access instruction and second access instruction, above-mentioned first access instruction be indication utilize above-mentioned prefetch buffer half come belonging to the instruction that above-mentioned first group memory bank carries out access, above-mentioned second access instruction be indication utilize above-mentioned prefetch buffer remaining half and with the shared above-mentioned prefetch buffer of above-mentioned first access instruction, come belonging to the instruction that above-mentioned second group memory bank carries out access.
According to this structure, the storer of having improved by use, and make that half the access of burst length of original basic burst length is paired, can improve the efficient that data transmit, wherein the storer of above-mentioned improvement have can be between two memory banks that belong to not on the same group shared prefetch buffer.
At this, above-mentioned storage control device also can be connected with a plurality of storeies that comprise above-mentioned storer, above-mentioned instruction generating unit also possesses the storer judging part, this storer judging part is according to judging that with above-mentioned access request physical address corresponding which in above-mentioned a plurality of storer be the data of wanting access belong to, according to the group judgement of judging part and the judgement of storer judging part, want the data of access to belong to 1 storer and stride two memory banks of the group that belongs to different and consecutive hours, above-mentioned instruction generating unit generates the above-mentioned a plurality of access instructions that comprise above-mentioned first access instruction and above-mentioned second access instruction.
According to this structure, can coming respectively according to storer and group, access can improve access efficiency from the instruction of main frame.
At this, S continuous on also can the line direction with above-mentioned view data pixel be as data block, and each data block belongs to and the different group of group that comprises with the data block of respective data blocks adjacency, and S is the integer more than or equal to 2.
According to this structure, can generate with the border equal number of data block to (first and second access instructions).
At this, above-mentioned data block also can be half size that is carried out the data of burst access with above-mentioned burst length N.
According to this structure, can get rid of non-paired access instruction, and only generate instruction (first and second access instruction), become the access of repetitive burst pulse length N/2 in fact, can improve data-transmission efficiency.
At this, above-mentioned data block also can be a minimum access unit.
According to this structure, can get rid of non-paired access instruction, and only generate instruction (first and second access instruction), become the access of repetitive burst pulse length N/2 in fact, can improve data-transmission efficiency.
At this, also can be that M the above-mentioned data block of adjacency on the column direction belongs to identical group, belong to and be different from the group that is included on the column direction with the group of other M data block of this M data block adjacency, M is the integer more than 2.
According to this structure, especially in the access of rectangular area, even from the row at place, access destination when other row shift, also can increase can be between two memory banks that belong to not on the same group the situation of shared prefetch buffer.
At this, also can be from capable by M on the row column direction of access by the row of access by above-mentioned first access instruction by above-mentioned second access instruction.At this, above-mentioned M also can be 2.
According to this structure, even be the field of rectangular area when reading, or frame can both increase the situation of shared prefetch buffer when reading.
In addition, accumulator system of the present invention has storer and above-mentioned storage control device, this memory stores view data, and be carried out the burst access by burst length N, N is the integer more than 2, above-mentioned storer possesses first group that comprises a plurality of memory banks, comprise second group of a plurality of memory banks, the prefetch buffer of N bit, above-mentioned prefetch buffer has first pattern and second pattern, this first pattern is from the look ahead data of N bit of a memory bank, respectively the look ahead data of N/2 bit of the memory bank that this second pattern is subordinated to different groups.
In addition, memory control methods of the present invention, be used to control access to storer, this storer possesses first group of comprising a plurality of memory banks, comprise second group of a plurality of memory banks, the prefetch buffer of N bit, wherein N is the integer more than 2, this memory stores view data also is carried out the burst access by burst length N, this memory control methods comprises: instruction generates step, access request according to the logical address that comprises the rectangular area in the above-mentioned view data of expression generates a plurality of access instructions that comprise physical address; And step is sent in instruction, send the above-mentioned a plurality of access instructions that generate by the instruction generating unit to above-mentioned storer, generate in the step in above-mentioned instruction, according to above-mentioned access request physical address corresponding, judge the data want access whether to stride two memory banks of the group that belongs to different and continuous, and generate above-mentioned first access instruction and above-mentioned second access instruction according to judged result, above-mentioned first access instruction be indication utilize above-mentioned prefetch buffer half come belonging to the instruction that above-mentioned first group memory bank carries out access, above-mentioned second access instruction be indication utilize above-mentioned prefetch buffer remaining half and with the shared above-mentioned prefetch buffer of above-mentioned first access instruction, come belonging to the instruction that above-mentioned second group memory bank carries out access.
Memory control methods of the present invention according to above-mentioned possesses: the instruction generating unit, and accept memory access requests, and generate access instruction at storer from main frame; The group judging part in above-mentioned instruction generating unit, judges to belong to which group by a plurality of groups that the are positioned at storer memory banks that constitute; Portion is sent in instruction, will send to storer at the memory instructions that above-mentioned instruction generating unit generates; And the numerical control control part, carry out the transmission of data according to the instruction sequences that the above-mentioned instruction portion of sending sends, the parallel access of carrying out at a plurality of groups.
According to this structure, data necessary is configured to different groups more equably, to little pixel data access the time, utilize half access of original basic burst length to come switch groups, carry out the high efficiency that data transmit thus.
The effect of invention:
In storage control device of the present invention and method, owing to needing higher frequency band to use under the situation of the DRAM with higher data movement capacity as entire system, use memory bank to be divided into a plurality of groups improvement DRAM, data necessary is configured in the different groups more equably, and switch groups carries out access, can realize with data movement capacity still less thus carrying out little pixel unit processing H.264 wait image processing techniques.Thus, even do not use further more high performance memory module, also can realize the system that data-transmission efficiency is high.
The application's technical background information
The application number that will propose on April 22nd, 2008 be in instructions, accompanying drawing and the claim scope of the Japanese patent application of 2008-111185 disclosed all the elements by with reference to covering among the application.
Description of drawings
Figure 1A is the figure of kind, operating frequency and burst length that DRAM of the prior art is shown.
Figure 1B is illustrated in the figure that is applicable to the memory bank interleaving access among the DRAM of the prior art and data configuration is shown.
Fig. 2 is illustrated in the figure that takes place in the data configuration of Figure 1B an example of the access of unwanted data.
Fig. 3 A is the block diagram of the configuration example of the improvement DRAM that comprised in the storer that illustrates in the embodiments of the present invention 1.
Fig. 3 B is the figure that the memory data configuration example of embodiments of the present invention 1 is shown.
Fig. 4 is illustrated in the access essential regions when in the data configuration of above-mentioned Fig. 3 B the access request identical with above-mentioned Fig. 2 having taken place.
Fig. 5 A is the block diagram that the structure of the storage control device in the embodiments of the present invention 1 is shown.
Fig. 5 B is the process flow diagram that the memory access control method of embodiments of the present invention 1 is shown.
Fig. 6 is the figure that the sequential legend in the memory control methods of the prior art is shown.
Fig. 7 is the figure that the sequential legend in the memory control methods of embodiments of the present invention 1 is shown.
Fig. 8 is the figure that the frame buffer structure in the memory control methods of embodiments of the present invention 1 is shown.
Fig. 9 is the figure that the image access method in the memory control methods of embodiments of the present invention 1 is shown.
Figure 10 is the figure that the frame buffer structure example in the memory control methods of embodiments of the present invention 1 is shown.
Figure 11 is the figure that the image access method in the memory control methods of embodiments of the present invention 1 is shown.
Figure 12 is the figure that the image access method in the memory control methods of embodiments of the present invention 1 is shown.
Figure 13 is the block diagram of structure that the storage control device of embodiments of the present invention 2 is shown.
Figure 14 is the process flow diagram that the memory control methods of embodiments of the present invention 2 is shown.
Figure 15 is the figure that the frame buffer structure example of embodiments of the present invention 2 is shown.
Figure 16 is the figure of the sequential legend the when image access of embodiments of the present invention 2 is shown.
Figure 17 illustrates the figure that is used to use system architecture example of the present invention.
Figure 18 illustrates the figure that has carried system LSI of the present invention and the application examples of system (set system) is set.
Description of drawings
101 storage control devices
102 instruction generating units
103 equipment judging parts
104 groups of judging parts
Portion is sent in 105 instructions
106 Data Control portions
107 main frames
108 address mapping portions
201DRAM
202 prefetch buffers
The 203P-S transformation component
Embodiment
Below, with reference to accompanying drawing embodiments of the present invention are described.
(embodiment 1)
The storage control device of embodiment 1 is controlled the access of storer, and this storer comprises: comprise first group of a plurality of memory banks; Comprise second group of a plurality of memory banks; N bit prefetch impact damper, N are the integer more than 2.This memory stores view data also is carried out the burst access by burst length N.
The structure example of the storer of storage control device institute access at first, is described.
Fig. 3 A is the block diagram that the structure example of the DRAM that comprises in the storer of storage control device institute access of embodiments of the present invention 1 is shown.In addition, data bus has 32 bits, 64 and manys the highway width of bit than top grade, but in the figure for the purpose of simplifying the description, only shows the part corresponding to 1 bit.For example, be under the situation of 32 bits at the data bus of storer, storer is so long as the structure that the DRAM of this figure is arranged side by side 32 is just passable.
The DRAM201 of this figure comprises: the first group of A that comprises 4 memory bank A0～A3; The second group of B that comprises 4 memory bank B0～B3; The prefetch buffer 202 of N (N is 8 among this figure) bit; And parallel serial conversion portion (following P-S transformation component) 203, be the DRAM201 (below be called improvement DRAM) that DDR3 has been carried out improvement.
Storer A0～A3, B0～B3 are equivalent to 1 memory cell array respectively, to the data of prefetch buffer 202 input and output by 1 bit of the memory cell array of row address and column address appointment.The operating frequency of each memory bank is that 133MHz is just passable.
Shown in dotted arrow and bracket among the figure, first pattern is the common access mode same with DDR3.Promptly, in first pattern, prefetch buffer 202 and by 1 memory bank of address appointment between with the operating frequency of the 133MHz data of input and output 8 bits concurrently, and via P-S transformation component 203 and data bus between with 8 times frequency 1066MHz inputoutput data serially.
Shown in solid arrow and bracket among the figure, second pattern is by belonging to 1 memory bank of first group and belonging to each shared half pattern of prefetch buffer 202 of 1 memory bank of second group.Promptly, in second pattern, prefetch buffer 202 in first group 1 memory bank and half (for example high-order 4 bits) of prefetch buffer 202 between with the operating frequency of 133MHz inputoutput data concurrently, simultaneously, between 1 memory bank in second group and prefetch buffer 202 remaining half (for example low level 4 bits) with the operating frequency of 133MHz inputoutput data concurrently.In addition, in second pattern, prefetch buffer 202 via P-S transformation component 203 and data bus between with 8 times the frequency 1066MHz data of input and output 8 bits serially.
Improvement DRAM is by first pattern action or by the action of second pattern, is that the access instruction (below be called memory instructions) by improvement DRAM being given via memory access control apparatus from main frame decides.Carry out first pattern according to 1 memory instructions.
Carry out second pattern according to two paired memory instructions (below be called first memory instruction and second memory instructs).
First memory instruction be indication utilize above-mentioned prefetch buffer half come belonging to the instruction that memory bank in above-mentioned first group carries out access.
Second memory instruction be indication utilize remaining half of above-mentioned prefetch buffer and with above-mentioned first memory instruct shared above-mentioned prefetch buffer, come the memory bank that belongs to second group is carried out the instruction of access.
Like this, improvement DRAM is by supporting second pattern, make two shared prefetch buffers 202 of memory instructions, (at this, the burst length N/2 of N=8) half (is 4 at this) carries out access with basic burst length N in these two memory instructions indications.Its result compares with first pattern of carrying out access with basic burst length N, in second pattern, cuts down useless data, can improve the service efficiency of data bus.In addition, the quantity of the group in the storer can be 2, also can be 3, just can be a plurality of so long as exclusively comprise the group of memory bank.In addition, the bank number in the group can be more than 1.
Fig. 3 B illustrates to have supposed in above-mentioned improvement DRAM especially with the DRAM that is made of two groups to be that an example makes the figure of an example effective situation of memory bank interleaving access, data configuration.Following alternate configurations data: be arranged in the 1st group, with the continuous data of same row address and with the data back of the N/2 byte of basic burst length * highway width/2 expressions, continue be arranged in second group, with the continuous data of same row address and with the data of the N/2 byte of basic burst length * highway width/2 expressions.In above-mentioned improvement DRAM because if group different then can with than basic access unit short send the transmission that covers other group, for example,, and can interrupt basic access unit with 1/2 if the group number is 2, then minimum access unit becomes basic access unit/2.
In addition, establishing the group number at this is 2, but constitutes also passable with the group number more than 2.
Fig. 4 is illustrated in the access essential regions when in the data configuration of above-mentioned Fig. 3 B the access request identical with above-mentioned Fig. 2 having taken place.Because minimum access unit is basic access unit/2,, can not produce unwanted data access so access request zone and access essential regions become identically as can be known.
Fig. 5 A is the block diagram of structure that the storage control device of embodiments of the present invention 1 is shown.The storage control device 101 of this figure comprises instruction generating unit 102, equipment judging part 103, Data Control portion 106.Instruction generating unit 102 possesses address mapping portion 108 and group judging part 104.
In Fig. 5 A, 107 pairs of storage control devices 101 of main frame send the instruction (below become host command) of storer being carried out access, the host command that 102 acceptance of instruction generating unit are sent by above-mentioned main frame 107, the instruction that generation is sent storer (below be called memory instructions), and send portion 105 to instruction and transmit memory instructions.Be arranged in the group judging part 104 of above-mentioned instruction generating unit 102, judging that whether according to the memory instructions that host command generated of above-mentioned main frame 107 be access at a plurality of groups certain group that is arranged in storer.If above-mentioned host command be storage at same unit area, then comprise to the access of same storer and to the access of a plurality of groups (for example organize A and organize B).Memory instructions in that the instruction generating unit 102 that comprises above-mentioned group of judging part 104 is generated sends above-mentioned instruction to after being divided by group and sends portion 105.Above-mentioned instruction is sent portion 105 at the memory instructions that is generated by above-mentioned instruction generating unit 102, and control is sent at ACTIVATE (activations), READ (reading), WRITE (writing), PRECHARGE (looking ahead) instruction of etc.ing of storer and controlled and send timing based on the memory instructions of the AC specification of storer etc.In Data Control portion 106, order information is sent in the instruction that acceptance is sent the memory instructions that portion 105 sends to storer 0 from above-mentioned instruction, at storer write access the time, accept data from main frame 107, and send order information according to instruction and be sent to storer 0, read from storer read access the time, accept data, and send order information according to instruction and transmit data to main frame 107 from storer 0.
Fig. 5 B is the process flow diagram that the memory access control method in the storage control device of the present invention is shown.The situation that host command is comprised the logical address of the rectangular area in the presentation video data describes.
Memory access control method is divided into two actions substantially.First is, the access request (host command) that instruction generating unit 102 bases comprise the logical address of the rectangular area in the presentation video data generates a plurality of access instructions (memory instructions) (step 51～58) that comprise physical address.Second is, portion 105 is sent in instruction and Data Control portion 106 will issue storer by a plurality of memory instructions that instruction generating unit 102 generates, and transmits by the data of access (step 57～step 60).
More specifically, address mapping portion 108 keeps the corresponding tables or the transformation rule of logical address and physical address in inside, if receive the above-mentioned host command that comprises logical address, will represent that then the logical address of rectangle is transformed to a plurality of physical addresss (step 51).
For example, represent in logical address under the situation of the rectangular area corresponding that address mapping portion 108 is transformed to 4 cover (set) physical addresss with Fig. 9 " data necessary ".That is, first cover becomes the physical address that has applied hatched part that points in the data block 0,8,4.Second cover becomes the physical address that has applied hatched part that points in the data block 1,9,5.The 3rd cover becomes the physical address that has applied hatched part that points in the data block 2,10,0.Quadruplet becomes the physical address that has applied hatched part that points in the data block 3,11,7.
Under the remaining situation of the cover of untreated physical address, instruction generating unit 102 is returned step 53, under the situation that the cover of untreated physical address not have to be left, enters step 57 (step 56).
Instruction is sent portion 105 corresponding to each memory instructions that is generated by instruction generating unit 102, comes the sending of each instruction (ACTIVATE, READ, WRITE, PRECHARGE etc.) at storer controlled and sent regularly and control.In Data Control portion 106, send the instruction that portion 105 accepts memory instructions from instruction and send order information, for at storer write access the time, accept data from main frame 107, and send order information according to instruction and transmit data to storer 0, to memory read access the time, accept data from storer 0, and send order information according to instruction and transmit data to main frame 107.
Fig. 6 as a comparative example, be among the Fig. 2 shown in the prior art read access request the time, the sequential legend on the memory bus.At this, the DRAM of institute's access is made as the DDR2 of the highway width with 32 bits, be that 4 situation is that example describes with basic burst length.As shown in Figure 2, be requested in centre in order to read the basic access unit of second memory bank, to send Read0 under the situation of N bit at t7 from basic access unit.Then, in order to read the basic access unit of first memory bank, send Read1 at t9.Data obtain following carrying out: in the timing of t10 and t11, obtain the reading of data of the basic access unit of second memory bank, in the timing of t12 and t13, obtain the reading of data of the basic access unit of first memory bank.But, be the data of the timing of t11 and t12 owing to be positioned at the data in access request zone, so just enough for the data of the timing of output t11 and t12 at the output data of main frame.As a result, active data is only at t11 and t12 in utilize the reading of data that obtains from t10 to t13, and transmission efficiency is 50%.
Fig. 7 be take place among Fig. 4 read access request the time, the sequential legend on the memory bus.At this, same with Fig. 6, be made as the DDR2 of highway width with DRAM with 32 bits with institute's access, be that 4 situation is that example describes with basic burst length.As shown in Figure 4, be requested in centre under the situation of N bit,, send Read0 at t7 in order to read second group basic access unit from basic access unit.Then, in above-mentioned improvement DRAM, if group is different, then can by than basic access unit short send the transmission that covers other group, so, send Read1 at t8 in order to read first group basic access unit.Data obtain following carrying out: in the timing of t10, obtain the reading of data of the basic access unit of second memory bank, in the timing of t11, obtain the reading of data of the basic access unit of first memory bank.Because access request zone and access essential regions are same, thus directly export at the output data of main frame t10 and t11 timing data just enough.As a result, active data is equally at t10 and t11, so transfer rate becomes 100% in the reading of data of utilizing t10 and t11 to obtain.
In addition, being made as 32 bits, basic burst width at this highway width with employed above-mentioned improvement DRAM is 4 to be illustrated, but being not limited to above-mentioned highway width and basic burst length, if basic burst length is more than 2, then can be any.
Fig. 8 illustrates the data configuration example when utilizing the data that are positioned on the above-mentioned improvement DRAM to constitute frame buffer.Frame buffer has 2-D data.Fig. 8 illustrates the collocation method of the data under the following situation: in the data at storer transmit, to carry out the data of access as the unit that is called data block with minimum access unit, use has the above-mentioned improvement DRAM of two groups A, B, has the system architecture of using 1 storer 0.At this, the unit area that will have different group A, a B in the horizontal direction with the order repeated configuration of " A, B, A, B " group A and group B, disposed same group of A or group B with band shape in vertical direction.
The example that Fig. 9 illustrates frame buffer at the data configuration method that has adopted above-mentioned Fig. 8 when carrying out the access of rectangular pixels data.When the pixel data that is illustrated as the rectangle of " data necessary " in to Fig. 9 carries out access, minimum access unit at DRAM determines, even so at the access of arbitrary data, carry out the data needs of access until carry out access till the border of data block as shown in Figure 9, the pixel data that the result becomes the rectangle that is expressed as " the actual data that are transmitted " carries out access.
Figure 10 is illustrated in data configuration when utilizing the data be arranged on the above-mentioned improvement DRAM to constitute frame buffer, is different from the other example of Fig. 8.Frame buffer has 2-D data.The collocation method of the data under the following situation has been shown: among Figure 10 in the data at storer transmit, same with Fig. 8, to carry out the data of access as the unit that is called data block with minimum access unit, use has the above-mentioned improvement DRAM of two groups A, B, has the system architecture of using 1 storer 0.At this, the unit area that will have different group A, a B in the horizontal direction with the order repeated configuration of " A, B, A, B " group A and group B, as " A, B, A, B ", disposed to clathrate group A in vertical direction or organized B.
The frame buffer that Figure 11 is illustrated at the data configuration method that has adopted above-mentioned Fig. 8 has the example that the actual data that are transmitted increase, data-transmission efficiency worsens under the situation of request of data shown in Figure 11.In above-mentioned improvement DRAM,, then can so for example can enough data blocks 8 come cover data piece 0, come cover data piece 1 with the transmission that sends other group of covering shorter than basic access unit with data block 9 if group is different.Equally, can cover 10, cover 11 with data block 3 with data block 2, but on the other hand, needing can cover data piece 4,5,6 and the transmission of 7 group B, so data block 12,13,14 and 15 also becomes the actual data that are transmitted, the result causes data-transmission efficiency to reduce.
In addition, at this,,, then can be the combination each other of arbitrary data piece so long as organize the combination each other of different data blocks though repeat to cover with the 0 such combination of data block 8 cover data pieces.
An example of the actual data that are transmitted when Figure 12 illustrates frame buffer at the data configuration method that has adopted above-mentioned Figure 10 the request of data identical with above-mentioned Figure 11 arranged.In above-mentioned improvement DRAM, if group is different, then can for example, can enough data blocks 1 come cover data piece 0 with the transmission that sends other group of covering shorter than basic access unit, come cover data piece 2 with data block 3.Equally, if cover 6 and transmit like that, then can more improve transmission efficiency in the highland than data configuration shown in Figure 8 by data block 9 coverings 8, data block 11 coverings 10, data block 5 coverings 4, data block 7.
In addition, at this,,, then can be the combination each other of arbitrary data piece so long as organize the combination each other of different data blocks though repeat to cover with the 0 such combination of data block 1 cover data piece.
In addition, the individual pixel of S (S is the integer more than 2) continuous on the line direction with the view data in the storer 0 is as data block, and each data block can belong to and be different from the group that comprises with the group of the data block of this data block adjacency.If like this, then can generate with the border equal number of data block to (first and second access instruction).
At this, above-mentioned data block also can be carried out half size of the data of burst access with above-mentioned burst length N.If like this, then can get rid of non-paired access instruction, only generate (first and second access instruction), become the access of repetitive burst pulse length N/2 in fact, can improve data-transmission efficiency.
In addition, above-mentioned data block also can be a minimum access unit.If like this, then can get rid of non-paired access instruction, only generate (first and second access instruction), become the access of repetitive burst pulse length N/2 in fact, can improve data-transmission efficiency.
At this, the individual above-mentioned data block of the M of adjacency on the column direction (M is the integer more than 2) belongs to identical group, also can belong to be included in column direction in the group different with the group of other M data block of this M data block adjacency.If like this, in the access of rectangular area particularly from the row at place, access destination when other row shift, also can be increased in the situation of shared prefetch buffer between two memory banks of the group that belongs to different.
At this, can be from capable by the row of access M to the column direction by the row of access by above-mentioned first access instruction by second access instruction.
In addition, M also can be 2.If like this,, and under the frame situation about reading, can both increase the situation of shared prefetch buffer even then under the situation about reading in the field of rectangular area.
In addition, accumulator system of the present invention comprises: storer, and storing image data, and be carried out burst with burst length N (N is the integer more than 2) and read; And above-mentioned storage control device.Above-mentioned storer has first group of comprising a plurality of memory banks, comprises second group and the prefetch buffer of N bit of a plurality of memory banks, above-mentioned prefetch buffer comprises first pattern and second pattern, this first pattern is from the look ahead data of N bit of a memory bank, respectively the look ahead data of N/2 bit of the memory bank that this second pattern is subordinated to different groups.
(embodiment 2)
In embodiments of the present invention 2, the inscape identical with embodiment 1 used identical Reference numeral, and omits explanation.
Figure 13 is the block diagram of structure that the storage control device of embodiments of the present invention 2 is shown.
In Figure 13,107 pairs of host computer control devices 101 of main frame send the instruction (below be called host command) of storer being carried out access, the host command that 102 acceptance of instruction generating unit are sent by above-mentioned main frame 107, the instruction that generation is sent storer (below be called memory instructions) is sent portion 105 to instruction and is transmitted memory instructions.Be arranged in the equipment judging part 103 of above-mentioned instruction generating unit 102, judgement is at storer 0 and 1 both sides' access or at wherein some accesses according to the memory instructions that host command generated of above-mentioned main frame 107, equally, be arranged in the group judging part 104 of above-mentioned instruction generating unit 102, judging that the memory instructions that host command generated according to above-mentioned main frame 107 is the access at a plurality of groups which group that is arranged in storer.If above-mentioned host command be storage at same unit area, then comprise to the access of same storer and to the access of a plurality of groups (for example organize A and organize B).The memory instructions that is generated in the instruction generating unit 102 that comprises the said equipment judging part 103 and above-mentioned group of judging part 104 is organized by each equipment and each and is sent to above-mentioned instruction after dividing and sends portion 105.105 pairs of memory instructions that generated by above-mentioned instruction generating unit 102 of portion are sent in above-mentioned instruction, and control is sent control and sent timing based on the memory instructions of the AC specification of storer etc. at ACTIVATE, READ, the instruction such as WRITE, PRECHARGE of storer.In numerical control control part 106, send the instruction that portion 105 accepts the memory instructions that sends to storer 0 and 1 from above-mentioned instruction and send order information, for at storer write access the time, accept data from main frame 107, and send order information according to instruction and send storer 0 and 1 to, read from storer read access the time, accept data from storer 0 and 1, and send order information according to instruction and transmit data to main frame 107.
Figure 14 is the process flow diagram of the memory control methods of embodiments of the present invention 2.
In Figure 14, in step 01, in the access request of above-mentioned instruction generating unit acceptance from above-mentioned main frame 107, the said equipment judging part 103 and above-mentioned group of judging part 104 judge it is which access in storer 0 and the storer 1, or to the access of which group in a plurality of groups, and generate with each storer and organize corresponding memory instructions.In step 02, send portion in above-mentioned instruction, judge that whether the memory instructions generated is the access to same unit area, be to enter step 03 under the situation to the access of same unit area, under no situation, enter step 05.Till following step 03 to 08, send in the portion in above-mentioned instruction and to carry out.In step 03, a plurality of storeies 0,1 are exported the memory instructions that has the common address of group A simultaneously, in step 04, storer 0,1 is exported simultaneously the memory instructions that has the common address of group B.In step 05, the group A corresponding address of output and storer 0, in step 06, the group A corresponding address of output and storer 1.In step 07, the group B corresponding address of output and storer 0, in step 08, the group B corresponding address of output and storer 1.In step 09, judge whether it is in above-mentioned Data Control portion, if, then enter step 10, otherwise (reading from storer) enters step 11 to the writing of storer at the writing of storer.In step 10, accept the data that transmit to storer from main frame 107, and send order information to storer 0 and storer 1 output data according to instruction.In step 11, accept data from storer 0 and storer 1, and send order information according to instruction and transmit data to main frame.
Memory instructions in the above-mentioned steps 01 generates processing, and roughly the step 51～step 56 with Fig. 5 B is identical, but following some difference.Promptly, according to the group judgement of judging part 104 and the judgement of equipment judging part 103, want the data of access to belong to 1 storer and stride two memory banks of the group that belongs to different and under the continuous situation, instruction generating unit 102 generates right that first memories instruction and second memory instruct.
Figure 15 illustrates the data configuration example of utilizing the data that are positioned on the storer to constitute the situation of frame buffer.Frame buffer has the data of two dimension, disposes A1 pixel data in the horizontal direction, has disposed A2 pixel data in vertical direction.Figure 15 illustrates the collocation method of the data under the following situation: in the data at storer transmit, will be with the data of minimum access size access as being called the such unit of data block, use has the DRAM of two groups A, B, has the system architecture of using two storeies 0,1.At this, the unit area that will have different group A, a B in the horizontal direction with the order repeated configuration of " A, B, B, A " group A and group B, respectively dispose 2 row group A and group B in vertical direction.To the pixel data of such frame buffer access rectangle the time, definite at the minimum access size of DRAM, so even at the access of arbitrary data, the data of carrying out access need until carry out access till the border of data block as shown in figure 15.
Figure 16 is a sequential legend when obtaining the pixel data of the rectangle among Figure 15, on the memory bus.When the pixel data to rectangle carries out access, in access, make CS become effective status simultaneously to storer 0 and storer 1 to unit area, simultaneously access is carried out in common address.In addition, at the access that is not unit area, make CS become effective status at storer 0 and storer 1 time of staggering, and different addresses is carried out access respectively.
According to above-mentioned structure, by possessing equipment judging part and the group judging part that is arranged in the instruction generating unit, can carry out access respectively according to equipment and group at instruction from main frame, can improve access efficiency.
In addition, in the present embodiment, will be arranged to the order line of storer 0 and storer 1 output from the instruction portion that sends shared, but also order line can be set independently respectively, also can only shared a part of order line, for example only high order bit of shared address or the only low-order bit of shared address.
In addition, about the data configuration in the frame buffer, not only can use the data configuration of present embodiment, can also adopt following data configuration:, longitudinal direction be disposed the data of same group of same memory bank with delegation the data block of transverse direction alternate configurations group A and the data block of group B.In this case, the column address of DRAM can be on transverse direction continuously, continuous with the data block of next line the column address that makes data block quantity is advanced on transverse direction after, also can skip 1 capable and continuous with data block every delegation.In addition, after the column address that makes data block quantity is advanced on transverse direction to every the data block configuration of delegation under the situation of continuous column address, same with present embodiment, also can take in image access, after data block is carried out access, to skip 1 the row and with the method for carrying out access every the data block of delegation.In this case, in the present embodiment, though be the data configuration that per two row become different groups, but owing to be the data of same memory bank with delegation, so instruct the mode of carrying out access to compare with common sending, also can carry out access so that the burst length of a plurality of data block quantity prolongs at each data block.
In addition, in the respective embodiments described above, illustrated that prefetch buffer 202 is belonging between two memory banks of two different groups by shared structure, but also can be shared between the memory bank more than 3 in the different group that belongs to more than 3.Under this situation, if the quantity of the memory bank of the prefetch buffer 202 of shared N bit is made as m, then as long as each shared memory bank respectively use the prefetch buffer 202 of N/m bit just passable.
In addition, among Figure 17,, show the system architecture example of recorded in blue system as using system architecture of the present invention.Figure 17 applies the present invention to the example of media with the memorizer control circuit among the LSI.The memorizer control circuit of this figure is corresponding to the storage control device of Fig. 5 A or Figure 13.In addition,, also can be applied to the regulating circuit in the dma control circuit, also can be applied to the regulating circuit in the cd playing control circuit though exemplified the memorizer control circuit of media with LSI as embodiment.
In addition, Figure 18 is the application examples of having carried system LSI of the present invention and system being set.The system LSI of this figure is equivalent to the media LSI among Figure 17.Such the present invention not only can be applied to system LSI, can also be applied in the various products such as pocket telephone, broadcast receiver, storing reproduction device, digital television, car-mounted terminal, automobile.
Use on the industry
Memory control methods of the present invention is as the memorizer control circuit in the system that carries out the image processing controls and useful. In addition, these also can be used in the image processing system of the digital AV appliance system such as television set or video recorder, recorder, video camera, pocket telephone or personal computer etc.
Claims (11)
1. storage control device, be used to control access to storer, this storer possesses first group of comprising a plurality of memory banks, comprise second group of a plurality of memory banks, the prefetch buffer of N bit, wherein N is the integer more than 2, this memory stores view data also is carried out the burst access by burst length N, it is characterized in that this storage control device comprises:
The instruction generating unit, the access request according to the logical address that comprises the rectangular area in the above-mentioned view data of expression generates a plurality of access instructions that comprise physical address; And
Portion is sent in instruction, sends the above-mentioned a plurality of access instructions that generated by the instruction generating unit to above-mentioned storer,
Above-mentioned instruction generating unit has the group judging part, this group judging part basis and above-mentioned access request physical address corresponding, judge which group the memory bank that comprises the data of wanting access belongs to, above-mentioned instruction generating unit is striden two memory banks of the group that belongs to different and consecutive hours in the data of wanting access, generation comprises above-mentioned a plurality of access instructions of first access instruction and second access instruction
Above-mentioned first access instruction be indication utilize above-mentioned prefetch buffer half come belonging to the instruction that above-mentioned first group memory bank carries out access,
Above-mentioned second access instruction be indication utilize above-mentioned prefetch buffer remaining half and with the shared above-mentioned prefetch buffer of above-mentioned first access instruction, come belonging to the instruction that above-mentioned second group memory bank carries out access.
2. storage control device according to claim 1 is characterized in that,
Above-mentioned storage control device is connected with a plurality of storeies that comprise above-mentioned storer,
Above-mentioned instruction generating unit also possesses the storer judging part, and this storer judging part basis judges that with above-mentioned access request physical address corresponding which in above-mentioned a plurality of storer be the data of wanting access belong to,
According to the group judgement of judging part and the judgement of storer judging part, want the data of access to belong to 1 storer and stride two memory banks of the group that belongs to different and consecutive hours, above-mentioned instruction generating unit generates the above-mentioned a plurality of access instructions that comprise above-mentioned first access instruction and above-mentioned second access instruction.
3. storage control device according to claim 1 and 2 is characterized in that,
S continuous on the line direction with above-mentioned view data pixel be as data block, and each data block belongs to respectively and the different group of group that comprises with the data block of respective data blocks adjacency, and S is the integer more than 2.
4. storage control device according to claim 3 is characterized in that,
Above-mentioned data block is half size that is carried out the data of burst access with above-mentioned burst length N.
5. storage control device according to claim 3 is characterized in that,
Above-mentioned data block is a minimum access unit.
6. storage control device according to claim 3 is characterized in that,
The M of adjacency above-mentioned data block belongs to identical group on the column direction, belongs to be different from the group that is included on the column direction with the group of other M data block of this M data block adjacency, and M is the integer more than 2.
7. storage control device according to claim 6 is characterized in that,
Is from capable by M on the row column direction of access by above-mentioned first access instruction by above-mentioned second access instruction by the row of access.
8. according to claim 6 or 7 described storage control devices, it is characterized in that,
Above-mentioned M is 2.
9. an accumulator system has storer and storage control device, and this memory stores view data also is carried out the burst access by burst length N, and N is the integer more than 2, it is characterized in that,
Above-mentioned storer possesses first group of comprising a plurality of memory banks, comprise second group of a plurality of memory banks, the prefetch buffer of N bit,
Above-mentioned prefetch buffer has first pattern and second pattern, and this first pattern is from the look ahead data of N bit of a memory bank, respectively the look ahead data of N/2 bit of the memory bank that this second pattern is subordinated to different groups,
Above-mentioned storage control device comprises:
The instruction generating unit, the access request according to the logical address that comprises the rectangular area in the above-mentioned view data of expression generates a plurality of access instructions that comprise physical address; And
Portion is sent in instruction, sends the above-mentioned a plurality of access instructions that generated by the instruction generating unit to above-mentioned storer,
Above-mentioned instruction generating unit has the group judging part, this group judging part basis and above-mentioned access request physical address corresponding, judge the data want access whether to stride two memory banks of the group that belongs to different and continuous, above-mentioned instruction generating unit generates the above-mentioned a plurality of access instructions that comprise first access instruction and second access instruction according to the group judgment result
Above-mentioned first access instruction be indication utilize above-mentioned prefetch buffer half come belonging to the instruction that above-mentioned first group memory bank carries out access,
Above-mentioned second access instruction be indication utilize above-mentioned prefetch buffer remaining half and with the shared above-mentioned prefetch buffer of above-mentioned first access instruction, come belonging to the instruction that above-mentioned second group memory bank carries out access.
10. a SIC (semiconductor integrated circuit) is characterized in that,
Formed each described storage control device in the claim 1～8.
11. memory control methods, be used to control access to storer, this storer possesses first group of comprising a plurality of memory banks, comprise second group of a plurality of memory banks, the prefetch buffer of N bit, wherein N is the integer more than 2, this memory stores view data also is carried out the burst access by burst length N, it is characterized in that this memory control methods comprises:
Instruction generates step, and the access request according to the logical address that comprises the rectangular area in the above-mentioned view data of expression generates a plurality of access instructions that comprise physical address; And
Step is sent in instruction, sends the above-mentioned a plurality of access instructions that generated by the instruction generating unit to above-mentioned storer,
Generate in the step in above-mentioned instruction, according to above-mentioned access request physical address corresponding, judge the data want access whether to stride two memory banks of the group that belongs to different and continuous, and generate above-mentioned first access instruction and above-mentioned second access instruction according to judged result
Above-mentioned first access instruction be indication utilize above-mentioned prefetch buffer half come belonging to the instruction that above-mentioned first group memory bank carries out access,
Above-mentioned second access instruction be indication utilize above-mentioned prefetch buffer remaining half and with the shared above-mentioned prefetch buffer of above-mentioned first access instruction, come belonging to the instruction that above-mentioned second group memory bank carries out access.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2008-111185 | 2008-04-22 | ||
| JP2008111185 | 2008-04-22 | ||
| PCT/JP2009/001815 WO2009130888A1 (en) | 2008-04-22 | 2009-04-21 | Memory controller, memory system, semiconductor integrated circuit, and memory control method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102016809A true CN102016809A (en) | 2011-04-13 |
Family
ID=41216629
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009801141803A Pending CN102016809A (en) | 2008-04-22 | 2009-04-21 | Memory controller, memory system, semiconductor integrated circuit, and memory control method |
Country Status (4)
| Country | Link |
|---|---|
| US (1) | US8918589B2 (en) |
| JP (1) | JP5351145B2 (en) |
| CN (1) | CN102016809A (en) |
| WO (1) | WO2009130888A1 (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8595459B2 (en) * | 2004-11-29 | 2013-11-26 | Rambus Inc. | Micro-threaded memory |
| JP5393405B2 (en) * | 2009-11-05 | 2014-01-22 | キヤノン株式会社 | Memory control circuit |
| JP2013089030A (en) * | 2011-10-18 | 2013-05-13 | Elpida Memory Inc | Information processing system, control system, and semiconductor device |
| JP6062714B2 (en) * | 2012-10-31 | 2017-01-18 | キヤノン株式会社 | MEMORY CONTROL DEVICE, MEMORY CONTROL METHOD, AND PROGRAM |
| US20160357690A1 (en) * | 2014-03-07 | 2016-12-08 | Mitsubishi Electric Corporation | Information processing device and information processing method |
| JP5911548B1 (en) * | 2014-10-23 | 2016-04-27 | インターナショナル・ビジネス・マシーンズ・コーポレーションＩｎｔｅｒｎａｔｉｏｎａｌ Ｂｕｓｉｎｅｓｓ Ｍａｃｈｉｎｅｓ Ｃｏｒｐｏｒａｔｉｏｎ | Apparatus, method, and computer program for scheduling access request to shared memory |
| US10067903B2 (en) | 2015-07-30 | 2018-09-04 | SK Hynix Inc. | Semiconductor device |
| US11755255B2 (en) * | 2014-10-28 | 2023-09-12 | SK Hynix Inc. | Memory device comprising a plurality of memories sharing a resistance for impedance matching |
| US20160232112A1 (en) * | 2015-02-06 | 2016-08-11 | Futurewei Technologies, Inc. | Unified Memory Bus and Method to Operate the Unified Memory Bus |
| US9990159B2 (en) * | 2015-06-26 | 2018-06-05 | Xitore, Inc. | Apparatus, system, and method of look-ahead address scheduling and autonomous broadcasting operation to non-volatile storage memory |
| KR102412609B1 (en) * | 2017-11-03 | 2022-06-23 | 삼성전자주식회사 | A memory device for storing and outputting an address accroding to an internal command and operating method thereof |
| JP2021039447A (en) | 2019-08-30 | 2021-03-11 | キヤノン株式会社 | Memory controller and method implemented by memory controller |
| US11137936B2 (en) | 2020-01-21 | 2021-10-05 | Google Llc | Data processing on memory controller |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1116763A (en) * | 1994-07-27 | 1996-02-14 | 株式会社日立制作所 | Semiconductor memory |
| JPH10144073A (en) * | 1996-11-08 | 1998-05-29 | Fujitsu Ltd | Access mechanism for synchronous dram |
| JPH11272550A (en) * | 1997-12-17 | 1999-10-08 | Fujitsu Ltd | Memory access method used for random access memory, memory access circuit, synchronous dynamic random access memory device and semiconductor memory |
| JP2002175689A (en) * | 2000-09-29 | 2002-06-21 | Mitsubishi Electric Corp | Semiconductor integrated circuit device |
| JP2005196485A (en) * | 2004-01-07 | 2005-07-21 | Matsushita Electric Ind Co Ltd | Dram control device and dram control method |
| CN1809817A (en) * | 2003-06-20 | 2006-07-26 | 飞思卡尔半导体公司 | Method and apparatus for dynamic prefetch buffer configuration and replacement |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH09190376A (en) | 1996-01-12 | 1997-07-22 | Oki Electric Ind Co Ltd | Memory controller |
| JP3732593B2 (en) | 1996-09-30 | 2006-01-05 | 株式会社東芝 | Image processing device |
| JP2000066950A (en) * | 1998-08-25 | 2000-03-03 | Toshiba Corp | Semiconductor storage device |
| JP3288327B2 (en) * | 1999-02-09 | 2002-06-04 | エヌイーシービューテクノロジー株式会社 | Video memory circuit |
| JP2000330864A (en) * | 1999-05-18 | 2000-11-30 | Fujitsu Ltd | Control of synchronous dram |
| US6795079B2 (en) * | 2001-02-15 | 2004-09-21 | Sony Corporation | Two-dimensional buffer pages |
| KR101377305B1 (en) * | 2005-06-24 | 2014-03-25 | 구글 인코포레이티드 | An integrated memory core and memory interface circuit |
| US7613883B2 (en) * | 2006-03-10 | 2009-11-03 | Rambus Inc. | Memory device with mode-selectable prefetch and clock-to-core timing |
| JP5055989B2 (en) * | 2006-12-08 | 2012-10-24 | 富士通セミコンダクター株式会社 | Memory controller |
-
2009
- 2009-04-21 JP JP2010509073A patent/JP5351145B2/en not_active Expired - Fee Related
- 2009-04-21 US US12/988,396 patent/US8918589B2/en not_active Expired - Fee Related
- 2009-04-21 CN CN2009801141803A patent/CN102016809A/en active Pending
- 2009-04-21 WO PCT/JP2009/001815 patent/WO2009130888A1/en active Application Filing
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1116763A (en) * | 1994-07-27 | 1996-02-14 | 株式会社日立制作所 | Semiconductor memory |
| JPH10144073A (en) * | 1996-11-08 | 1998-05-29 | Fujitsu Ltd | Access mechanism for synchronous dram |
| JPH11272550A (en) * | 1997-12-17 | 1999-10-08 | Fujitsu Ltd | Memory access method used for random access memory, memory access circuit, synchronous dynamic random access memory device and semiconductor memory |
| JP2002175689A (en) * | 2000-09-29 | 2002-06-21 | Mitsubishi Electric Corp | Semiconductor integrated circuit device |
| CN1809817A (en) * | 2003-06-20 | 2006-07-26 | 飞思卡尔半导体公司 | Method and apparatus for dynamic prefetch buffer configuration and replacement |
| JP2005196485A (en) * | 2004-01-07 | 2005-07-21 | Matsushita Electric Ind Co Ltd | Dram control device and dram control method |
Also Published As
| Publication number | Publication date |
|---|---|
| JP5351145B2 (en) | 2013-11-27 |
| WO2009130888A1 (en) | 2009-10-29 |
| JPWO2009130888A1 (en) | 2011-08-11 |
| US20110035559A1 (en) | 2011-02-10 |
| US8918589B2 (en) | 2014-12-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102016809A (en) | Memory controller, memory system, semiconductor integrated circuit, and memory control method | |
| EP1474747B1 (en) | Address space, bus system, memory controller and device system | |
| US7779215B2 (en) | Method and related apparatus for accessing memory | |
| US6553449B1 (en) | System and method for providing concurrent row and column commands | |
| CN101495975B (en) | Memory control device, memory device, and memory control method | |
| US7814294B2 (en) | Memory device, memory controller and memory system | |
| US8661180B2 (en) | Memory controlling device and memory controlling method | |
| JP2008544424A (en) | System and method for improving parallel processing of DRAM | |
| US20210280226A1 (en) | Memory component with adjustable core-to-interface data rate ratio | |
| JP6159478B2 (en) | Data writing method and memory system | |
| EP2092759B1 (en) | System for interleaved storage of video data | |
| KR101086417B1 (en) | Apparatus and method for partial access of dynamic random access memory | |
| CN100446084C (en) | Picture data transmitting method, video data transmitting method and time-sequence control module | |
| CN102541769B (en) | Memory interface access control method and device | |
| CN100444636C (en) | Method for improving SDRAM bus efficiency in video decoder | |
| CN100536021C (en) | High-capacity cache memory | |
| CN114442908B (en) | Hardware acceleration system and chip for data processing | |
| US8806132B2 (en) | Information processing device, memory access control device, and address generation method thereof | |
| US8347026B2 (en) | Memory device and memory device control method | |
| US8581918B2 (en) | Method and system for efficiently organizing data in memory | |
| WO2023189358A1 (en) | Memory control device | |
| CN102073604A (en) | Method, device and system for controlling read and write of synchronous dynamic memory | |
| CA2802666A1 (en) | Image processing device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20110413 |